CN103248895A - Quick mode estimation method used for HEVC intra-frame coding - Google Patents

Abstract

The invention provides a quick mode estimation method used for HEVC (High Efficient Video Coding) intra-frame coding. In the specification of the HEVC video coding, the available modes of the infra-frame coding are more complex; and an optimal mode is chosen from more forecasting angles during coding, so that the arithmetic complexity of an encoder is greatly increased. The invention provides the method which comprises the following steps: firstly, calculating integral point forecasting errors of every row or line and mode angles corresponding to forecasting errors; then estimating the forecasting errors of every row or line at the angles corresponding to various modes through the interpolation method; and finally, combing the forecasting errors of the corresponding modes of every row or line to serve as the forecasting error of a whole forecasting mode. Through the adoption of the method, estimations for forecasting errors in intra-frame prediction at different angles can be finished more quickly, so as to be helpful to make a final decision of a better forecasting mode.

Description

A kind of quick mode method of estimation for the HEVC intraframe coding

Technical field

The invention belongs to image processing field, particularly for reducing the computational complexity of predicting that when carrying out the intraframe coding of HEVC (High Efficient Video Coding) efficient video coding angle mode is estimated.

Background technology

In April, 2010, two big international video encoding standards organize VCEG and MPEG to set up the video compression associating JCT-VC of group (Joint collaborative Team on Video Coding), together develop the HEVC standard, and H.265 it be also referred to as.HEVC standard main target is H.264/AVC to realize the significantly raising of code efficiency with the previous generation standard, especially at high-resolution video sequence.Its target be same video quality (PSNR) down code check reduce to 50% of standard H.264.

With regard to the current stage, HEVC still continues to use and H.264 just begins the hybrid encoding frame that adopts.Interframe and intraframe predictive coding: the correlation of eliminating time-domain and spatial domain.Transition coding: residual error is carried out transition coding to eliminate spatial coherence.Entropy coding: eliminate statistical redundancy.HEVC will put forth effort to study new coding tools or technology in hybrid encoding frame, improve video compression efficiency.

At present, the new features of the many codings that proposed in the discussion of JCT-VC tissue might add in the HEVC standard, the concrete document of each time discussion can from Http:// wftp3.itu.intObtain.

Big young pathbreaker's maximum of encoding block can reach 64x64, and this mainly is the application for the HD video compressed encoding.The dct transform coding will be broken through 8x8, and maximum can reach 32x32.For infra-frame prediction, the refinement more of the direction of prediction reaches 35 kinds of infra-frame predictions (intra predication) pattern, and Fig. 2 shows the situation of 35 kinds of intra prediction modes, and this will make infra-frame prediction better accurately, reduce redundant more efficiently.Inter prediction has adopted more multitap filter when interpolation, and 1/4 pixel precision, improves the precision of inter prediction.At the entropy encoding context, used the stronger CABAC of adaptability.

Aspect infra-frame prediction, owing to used nearly 35 kinds of intra prediction modes, therefore, when encoding in the conducting frame, if directly at 35 kinds of intra prediction mode utilization rate aberration optimizing (Rate-Distortion optimization, RDO) judge optimum prediction mode, then encoder can not bear the operand that all prediction direction is carried out RDO.Operand when carrying out the RDO computing in order to reduce encoder, HM4.0 (HEVC test model4.0) has adopted the method for Hadamard conversion, and (Rough Mode Decision RMD) selects several better candidate patterns by a kind of coarse mode decision-making.Subsequently, obtaining optimal mode at carrying out the RDO computing by these several preferable candidate pattern.

In the JCTVC-C207 motion that JCT-VC has adopted that the residual signals Hadamard transform coefficients is absolute with (minimum absolute sum of Hadamard Transformed coefficients of residual signal HSAD) counts computational process with mode bit and adds coarse mode decision-making (RMD) algorithm.Among the motion JCTVC-D283, to unite the direction intraframe prediction algorithm and simplify, and add extraction maximum probability pattern (most probable mode, MPM) algorithm, namely after RMD, use the MPM algorithm, thereby guarantee in the candidate pattern set of carrying out the RDO computing, to comprise MPM all the time.Motion JCTVC-C218 simplifies the redundant process that CU is divided into PU, and these shortcut calculations have reduced the complexity of infra-frame prediction, but these schemes all are the most basic solutions and do not consider the character of piece.

Aspect infra-frame prediction, though proposed some kinds of candidate pattern selection algorithms, reduce the pattern quantity of finally carrying out RDO, the quantity of candidate pattern still has the space of decline.

HEVC (High Efficient Video Coding) provides higher compliant compression to finish compression to video, particularly at the compression of large-size images, provides more effective instrument.For intraframe coding, HEVC has not only proposed new predictive mode (planar prediction), also original 8 prediction direction is H.264 increased to 33 prediction direction.As shown in Figure 1, except Planar (planar prediction) mode and DC mode, there are 33 prediction direction available.Be example as Fig. 2 with pattern 29, the angle forecast method has been described.Namely the value of the encoded pixel by block to be encoded top and left side is predicted the value of the pixel in the piece.More prediction direction has increased accuracy for predicting, but has also increased the complexity of coding simultaneously.Requirement according to standard predicts that to more orientation angle selecting best predictive mode is the problem that HEVC intraframe coding unit must solve.

By the HEVC standard as can be known, the problem that infra-frame prediction need solve mainly is two aspects: the firstth, the coding unit (as 64x64) of maximum is effectively divided, determine with great as coding unit, the secondth, after the size of coding unit is determined, select great and which kind of pattern as the predictive mode of present encoding unit.In the identifying code of the encoder that HEVC provides, the solution of above-mentioned two problems is solved in reverse order.At first all possible coding unit size is all carried out by thick to smart prediction estimation and mode decision, thereby calculate under the every kind of size of encoding various optimum cost.Again with the coding cost of the coding cost addition of fritter and big coding unit relatively.Which kind of size is total coding cost minimum of which kind of coding size then select for use as the size of coding unit.

In H.264 encoding, at angle how to predict intraframe coding, many algorithms have just been proposed.The prediction angle is estimated also have by the mode of progressively search and seek the local optimum point such as coming by the method for frequency domain or gradient.These methods also are applied in the intraframe coding prediction of HEVC.But at the new characteristics among the HEVC, some study also underway more targetedly.

Summary of the invention

In order to solve the technical problem of above-mentioned HEVC frame mode prediction, the present invention proposes a kind of new efficient video coding (HEVC) intra-frame encoding mode forecast method that is used for, comprise following steps:

The first step utilizes the coding result on top or the left side as predicting the outcome pixel value, directly obtains the prediction error value of the integer pixel of every row or column with image block computing to be encoded;

Second step, the angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance;

In the 3rd step, come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition by the integer pixel prediction error value of every row or column and the angle value of correspondence thereof;

In the 4th step, the prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode;

The 5th step, predicated error under the different mode that estimates with the method is as the coding cost of assessing different mode with this prediction on intra-frame prediction mode error, thereby find out optimum predictive mode or find out several more excellent predictive modes, for further mode decision provides little scope of assessment.

According to an aspect of the present invention, in the first step, directly utilize the direct conduct of coding result on top or the left side to predict the outcome, rather than adopt filtered coding result as predicting the outcome.

According to an aspect of the present invention, in the 4th step, the predicated error calculating of whole predictive mode is broken the whole up into parts, merge the predicated error of the corresponding pattern of conduct by the predicated error of estimating every row or column correspondence.

According to an aspect of the present invention, in the first step, the coding result that utilizes top or the left side is as predicting the outcome pixel value, and directly the prediction error value that obtains the integer pixel of every row or column with image block computing to be encoded is (be example with the 4x4 piece) realized by following formula:

$\mathrm{SAD}\left(\mathrm{mode}\right)=\underset{x=0..3}{\mathrm{\Σ}}\underset{y=-..3}{\mathrm{\Σ}}|S(x,y)-P(\mathrm{mide},x,y)|$

Wherein, S (x y) is the coded image gray value, mode be 0 (Planar pattern) or 1 (DC pattern) or 2 to 34, P (mode, x, y) for according to predicting the outcome that different mode calculates, x=0..3, y=0..3.

According to an aspect of the present invention, in the 4th step, the prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges realizes (being example with the lateral prediction) as the prediction error value of associative mode by following formula:

$\mathrm{SAD}\left(\mathrm{mode}\right)=\underset{x=0..3}{\mathrm{\Σ}}\mathrm{SAD}\_\mathrm{Interploted}(\mathrm{mod}e,x),\mathrm{mode}=19..34$

Wherein, and SAD_Interploted (mode, x)=(Int_SAD (mode_0, x) * (mode_1-mode)+Int_SAD (mode_1, x) * (mode-mode_0))/(mode_1-mode0), mode=19..34

Wherein mode_0 be in the corresponding x value less than the Mode value that equals current calculating mode, mode_1 is greater than the Mode value of the current mode of calculating in the corresponding x value.

According to an aspect of the present invention, be unit with limit (horizontal edge or longitudinal edge), calculate the corresponding reference edge scope of prediction angle.And only in this scope, only select for use integral point to carry out prediction error value calculating as calculating predicted value.

According to an aspect of the present invention, the prediction error value to each limit calculates corresponding angle value, i.e. Dui Ying mode value.And 33 (angle, prediction error values) being proposed by interpolation method estimation HEVC by these (angle, prediction error values).

According to an aspect of the present invention, the prediction error value under equal angular (pattern) on every limit is synthetic, obtain the prediction error value of total fritter, i.e. the prediction error value of each angle predictive mode.And with these prediction error values cost estimation of encoding, finish or the auxiliary decision-making of finishing frame mode.

The invention allows for a kind of device for the prediction of efficient video coding (HEVC) intra-frame encoding mode, comprising: infra-frame prediction is selected module, is used for intra-frame encoding mode is predicted selection, and the prediction selection result sends to intra-framed prediction module; Intra-framed prediction module is used for intra-frame encoding mode is predicted, and prediction result output is used for the predictive mode that infra-frame prediction selection module is selected the best.

According to an aspect of the present invention, described infra-frame prediction selects module also to be used for realizing: utilize the coding result on top or the left side as predicting the outcome pixel value, directly obtain the prediction error value of the integer pixel of every row or column with image block computing to be encoded; The angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance; Integer pixel prediction error value and corresponding angle value thereof by every row or column come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition; The prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode.

According to an aspect of the present invention, described intra-framed prediction module also is used for realizing: the predicated error under the different mode that estimates is assessed the coding cost of different mode as this prediction on intra-frame prediction mode error, thereby finds out optimum predictive mode or find out several predictive modes than top grade.

Description of drawings

The present invention is further described in more detail below in conjunction with drawings and the specific embodiments:

Accompanying drawing 1 is 35 kinds of definition schematic diagrames of intra prediction mode in the HEVC standard.

Accompanying drawing 2 is example explanation frame mode forecast method with pattern 29.

In the accompanying drawing 3 explanation HEVC reference encoder devices at the intra prediction mode decision-making technique flow chart of the coding unit of a certain size.

Accompanying drawing 4 is the intra prediction mode decision-making technique flow chart of coding unit at a certain size proposed by the invention.

Accompanying drawing 5 is the encoder block diagram according to the HEVC of the embodiment of the invention.

Embodiment

In order to make purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explaining the present invention, and be not used in restriction the present invention.

The method of estimation infra-frame prediction angle proposed by the invention is finished at the new feature of HEVC.This algorithm is at first by only estimating the prediction error value of intraframe prediction algorithm gained to the infra-frame prediction error amount of the integral point position of every row or column, obtain the predicated error of the defined various patterns of standard of every row or column then by the method for interpolation, at last it is synthesized the predicated error of predictive mode of whole correspondence.This algorithm has reduced computational complexity under the prerequisite of not losing forecasting accuracy.

Under the predictive mode of different prediction angle correspondences, namely predict when angle changes, near the selected prediction reference pixel variation of the pixel of edge (left side and top) not quite, the selected prediction reference pixel of pixel away from edge (left side and top) then has bigger variation.In the algorithm of the present invention, only used less reference point when calculating near the predicated error on prediction reference limit, then adopted more reference point when calculating away from the predicated error on prediction reference limit, thereby reach the active balance of operation effect and operand.In addition, in the HEVC algorithm, prediction is to calculate gained by the pixel on acquired top or the left side according to the different some interpolation of the different choice of prediction angle, then directly utilizes the prediction reference limit as predicting the outcome among the present invention, has reduced the computing of interpolation.

Fig. 5 shows the general configuration figure of the video encoder that efficient video coding (HEVC) realizes.The encoder architecture of HEVC is with H.264 employed encoder architecture is roughly the same, mainly be to have carried out further research, improvement at employed algorithm in each module, especially at high-resolution video sequence, its improved target be same video quality (PSNR) down code check reduce to 50% of standard H.264.

Because the encoder architecture of HEVC with H.264 employed encoder architecture is roughly the same, therefore in order not obscure the present invention, is only paid close attention to for the intra prediction mode of HEVC and is selected and infra-frame prediction.

Infra-frame prediction is a kind of new intraframe coding method that H.264/AVC at first adopts in the video encoding standard, and purpose is the spatial redundancy of removing between current block and the adjacent blocks.In infra-frame prediction, prediction piece P is formed by the piece after previous coding in the anter (slice) and reconstruction.Current block deducts this prediction piece P, produce residual block, residual block produces the conversion coefficient X behind a kind of the quantification through piece conversion, quantification back, again through the entropy coding, other information required with decoder-side (for example, predictive mode quantization parameter, motion vector etc.) are formed compression bit stream together and are used for transmission and decoding.In video encoding standard H.264/AVC, at the piece of 4*4,9 kinds of predictive modes that comprise direct current (DC) are arranged, at the piece of 8*8 and 16*16, have to comprise DC and plane (planar) in 4 kinds of interior patterns.When the conducting frame inner estimation mode is selected, from the predictive mode of supporting, choose the pattern that has optimum Match (for example, having minimum rate distortion (RD)) with current block.

In HEVC, adopted size to be respectively 4*4 at infra-frame prediction, 8*8,16*16, the prediction piece of 32*32 and 64*64 (predication unit), and possible predictive mode is shown in Figure 2.As shown in Figure 2, nearly 35 kinds predictive mode comprises 33 kinds of angle intra prediction mode directions, and DC (direct current) pattern and Planar (plane) pattern.

Number of times for the RD computing that reduces the infra-frame prediction complexity, the intra prediction mode of HEVC is used coarse mode decision-making (Rough Mode Decision, RMD) and extract MPM (most probable mode, MPM) method has obtained the optimal candidate set of modes of simplifying, thereby reduces the RD operation times.The RMD purpose is to guess the optimum frame inner estimation mode roughly, and its selection course is considered HSAD operation values and mode bit number, but RMD does not excavate the correlation between the extraction pattern.Therefore, the rule between the analysis extraction pattern can further reduce the complexity of associating infra-frame prediction.

The RMD algorithm

Operand when carrying out the RDO computing in order to reduce encoder, HM4.0 has at first adopted the method for Hadamard conversion, selects several better candidate patterns by RMD.Subsequently, obtaining optimal mode at carrying out the RDO computing by these several preferable candidate pattern.

Particularly, in RMD, use following formula to come to obtain 8 candidate pattern at the prediction piece of 4*4 and 8*8 size, obtain 3 candidate pattern at the prediction piece of 16*16 and 32*32 size:

$J=\mathrm{HSAD}\left(\mathrm{Re}{s}_{\mathrm{mode}}\right)+\mathrm{\λ}\·B_{\mathrm{mode}},$

Wherein, J is cost, _ModeBe all intra prediction modes of 35 kinds of being up to of HEVC regulation, HSAD (Res _Mode) expression residual signals Hadamard transform coefficients absolute and, B _ModeThe consumption figures of expression mode bit number, λ represents Lagrangian constant.

Therefore, the RMD algorithm just is to use above-mentioned formula in whole predictive modes of supporting (17 or 35), chooses 3 (at 16*16 and 32*32 size) or 8 (at 4*4 and 8*8 size) candidate pattern with minimum cost.

More specifically, in the RMD algorithm, at each pattern in whole predictive modes of supporting, the cost J that calculating is associated, and according to the descending of the cost J that is associated each predictive mode is sorted, and choose preceding 3 (at 16*16 and 32*32 sizes) or 8 (at 4*4 and 8*8 size) predictive modes as candidate pattern, add in the candidate pattern set.

Because the complexity of above-mentioned formula is more much smaller than the complexity of RDO in realization, therefore using before the RDO algorithm determines optimum prediction mode, using the RMD algorithm in whole predictive modes of supporting, to carry out once preselectedly can reducing the infra-frame prediction time significantly.

The RDO algorithm

RDO is the algorithm that generally uses in infra-frame prediction and the inter prediction in video coding, and it determines best matching blocks based on least mean-square error and (SSE:Sum of Square Error) usually.

The MPM algorithm

The MPM algorithm is the algorithm that has just proposed in HM3.0, and is further improved in JCTVC-D283.

Determine that to MPM in the various motions that relate to the MPM algorithm variety of way is arranged.This paper has used the MPM algorithm among the JCTVC-D283.Particularly, after the optimal mode candidate collection of using after the RMD algorithm has obtained to simplify, with the optimal mode of the optimal mode of the left side of current prediction piece prediction piece and top prediction piece directly as two maximum probability patterns (MPM), and judge whether MPM is included in this candidate collection, if be not included in this set, then add this two MPM.In other words, regardless of the model selection algorithm before the RDO, the optimal mode of piece and the optimal mode of top prediction piece are predicted in the left side that all adds current prediction piece.

In HM4.0, the associating infra-frame prediction extracts intra prediction mode and had for two steps.The first step, optimal mode candidate collection (RMD) is selected in all pattern process HSAD computings.In this optimal mode candidate collection, number of modes is by the different youngster's differences of PU size.Comprise 8 candidate pattern in 4*4 and the set of 8*8 sized blocks, and comprise 3 in 16*16 and the set of 32*32 sized blocks.Then, the MPM that extracts from contiguous block also adds in the set.Second step, each pattern in the optimal mode set optimized through RD calculate (RDO), its RD consumption figures minimum carried out infra-frame prediction as optimal mode.

According to one embodiment of present invention, for every kind of possible coding unit size, its mode decision process as shown in Figure 3.After HEVC frame mode prediction beginning, at first calculate the rough RD value of 34 kinds of patterns, finish the rough estimate to frame mode; Then select the top n pattern according to the rough estimate result and add the alternative mode set; Find out the predictive mode set of maximum possible according to the pattern of peripheral piece; Above-mentioned two set are merged as rate-distortion optimization (Rate-Distortion optimization, RDO) estimation model set; Adopt full RDO mode decision to select optimization model; Finish the prediction of HEVC frame mode.During wherein pattern rough estimate was calculated, identifying code had calculated it respectively at 33 directions and has predicted the outcome, and carried out with original picture block to be encoded that computing obtains corresponding predicated error and the cost of encoding.

Intra prediction mode comprises two classes among the HEVC: angle mode (pattern 2 to 34) and non-angle mode (planar pattern 0 and DC pattern 1).In angle mode, are linear change from the prediction angle of pattern 2 to 34 correspondences, therefore can release corresponding predictive mode according to the angle of prediction.But angle is continuous, and pattern disperses.And because only calculated the integer pixel point herein, corresponding pattern has just become " mark " value.It is not the pattern in the standard.

According to one embodiment of present invention, also proposed a kind of for HEVC intraframe coding angle mode forecast method, the method includes the steps of: the first step, the coding result that utilizes top or the left side directly obtain the prediction error value of the integer pixel of every row or column with image block computing to be encoded as the pixel value that predicts the outcome; Second step, the intra prediction mode of each row or column of the angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance and correspondence thereof; In the 3rd step, come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition by integer pixel prediction error value and the corresponding mode value thereof of every row or column; In the 4th step, the prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode; The 5th step, predicated error under the different mode that estimates with the method is as assessing the coding cost of different mode with this prediction on intra-frame prediction mode error, provides little scope of assessment thereby find out optimum predictive mode or find out several predictive modes than top grade for further mode decision.

In this embodiment, in the first step, the coding result that can also directly utilize top or the left side directly as predict the outcome rather than adopt its according to the different interpolation result of angle as predicting the outcome.

In second step, can also utilize near the corresponding integer pixel prediction of the column or row of the left side or top and have only less possibility, and with adjacent limit bigger correlation is arranged, reduced operand and do not have or less reduce the accuracy that predicated error is calculated.

In second step went on foot to the 4th, the predicated error of whole predictive mode can also be calculated and break the whole up into parts, merge the predicated error of the corresponding pattern of conduct by the predicated error of estimating every row or column correspondence.

Present embodiment be used for HEVC intraframe coding angle mode decision-making method of operation, not only can be used for the rough calculation to intra prediction mode, can also cooperate effective code check evaluation method, be directly used in the decision-making of final intra prediction mode.

The specific implementation of the described method of present embodiment can be referring to accompanying drawing 4, and its specific implementation process comprises following steps: after the intra prediction mode decision-making technique begins, computation schema 0(planar prediction planar at first) and pattern 1(DC) prediction error value; According to the horizontal edge integer predicted value of calculating gained in advance every horizontal edge being carried out the integer predicated error calculates; Calculate the prediction error value of the associative mode 2-pattern 18 of every horizontal edge by interpolation; The prediction error value of each horizontal edge associative mode 2-pattern 18 is merged the acquisition model prediction error value; According to the vertical edge integer predicted value of calculating gained in advance every vertical edge being carried out the integer predicated error calculates; Calculate the prediction error value of the associative mode 19-mode 34 of every vertical edge by interpolation; The prediction error value of each vertical edge associative mode 19-mode 34 is merged acquisition model rough estimate prediction error value; Count estimated value according to the prediction error value of 35 kinds of patterns and the bit of corresponding different mode, calculate total cost and select optimum candidate pattern by the RD method; End frame inner estimation mode decision-making technique.

Set forth for convenient, present embodiment is example with the infra-frame prediction of 4x4 piece.In to the calculating of predicated error, be that example is calculated with SAD (absolute error and).

1. calculating the prediction error value of Planar pattern (0) and DC pattern (1), is example with SAD:

Calculate the P that predicts the outcome (mode, x, y), x=0..3, y=0..3; In the following formula, p[x, y] be the coding result of the piece around the current block, if on every side piece as yet coding finish, then use the gray value of the point behind its nearest coding.With corresponding calculating in the HEVC standard, only be that 4x4 has been done corresponding simplification herein.Detail formula please refer to the corresponding chapters and sections of HEVC standard.

To the Planar pattern:

P(0,x,y)=((3-x)*p[-1,y]+(x+1)*p[4,-1]+(3-y)*p[x,-1]+(y+1)*p[-1,4]+4)>>(Log2(4)+1),x=0..3,y=0..3

To the DC pattern, because the image conversion in the coding generally is from the spatial domain to the spatial frequency domain, so the average gray of pixel is called DC coefficient, i.e. DC value, and other coefficient is called ac coefficient, i.e. AC value.

At first calculate the DC value, i.e. DCVal,

$\mathrm{DCVal}=\left(\underset{x^{\text{'}}=0..3}{\mathrm{\Σ}}p\right[x^{\text{'}},-1]+\underset{y^{\text{'}}=0..3}{\mathrm{\Σ}}p[-1,y^{\text{'}}]+4)>>(\mathrm{Log}2\left(4\right)+1)$

Calculate each predicted value (to brightness) then:

P(1,0,0)=(p[-1.0]+2*DCVal+p[0,-1]+2)>>2

P(1,x,0)=(p[x,-1]+3*DCVal+3)>>2,x=1..3

P(1,0,y)=(p[-1,y]+3*DCVal+2)>>2,y=1..3

P(1,x,y)=DCVal,x=1..3,y=1..3

Calculate its corresponding SAD then.Can be calculated as follows:

$\mathrm{SAD}\left(\mathrm{mode}\right)=\underset{x=0..3}{\mathrm{\Σ}}\underset{y=-..3}{\mathrm{\Σ}}|S(x,y)-P(\mathrm{mode},x,y)|$

Wherein (x y) is the coded image gray value to S, and mode is 0 (Planar pattern) or 1 (DC pattern).

2. vertically predict the SAD estimation (pattern 2 to 18) of angle

2.1 carry out integer SAD prediction when each horizontal edge is done vertical prediction;

$\mathrm{Int}\_\mathrm{SAD}(\mathrm{mode},y)=\underset{x=0..3}{\mathrm{\Σ}}|S(x,y)-\mathrm{Top}(x+\mathrm{ModeOffset}\left(\mathrm{mode}\right))|$

Wherein, y=0..3, (x y) is the coded image gray value to S, and ModeOffset (mode) is the side-play amount of corresponding different mode, and Top (x) is the image prediction value of prediction piece top delegation.Different mode (Mode) and corresponding ModeOffset[to different y values] be defined as follows (be example with the 4x4 piece).

Y	Mode	ModeOffset(mode)
			0	2,10,18	-1,0,1
1	2,6,10,14,18	-2,-1,0,1,2
			2	2,4.67,7.33,10,12.67,15.33,18	-3,-2,-1,0,1,2,3
3	2,4,6,8,10,12,14,16,18	-4,-3,-2,-1,0,1,2,3,4

2.2 (mode, y) interpolation is estimated the sad value of each horizontal edge associative mode 2 to 18 with Int_SAD.This sentences linear interpolation is the sad value that example is calculated each pattern.

SAD_Interploted(mode,y)=(Int_SAD(mode_0,y)*(mode_1-mode)+Int_SAD(mode_1)*(mode-mode_0))/(mode_1-mode_0)Mode=2..18

Wherein mode_0 be in the last table in the corresponding y value less than the Mode value that equals current calculating mode, mode_1 be in the last table in the corresponding y value greater than the Mode value of the current mode of calculating.

2.3 the sad value that computation schema is total;

$\mathrm{SAD}\left(\mathrm{mode}\right)=\underset{y=0..3}{\mathrm{\Σ}}\mathrm{SAD}\_\mathrm{Interploted}(\mathrm{mode},y),\mathrm{mode}=2..18$

Wherein (mode y) is step 2.2 interpolation gained to SAD_Interploted.

3. lateral prediction SAD estimation (pattern 19 to 34)

3.1 each longitudinal edge is carried out integer SAD prediction;

$(\mathrm{mode},x)=\underset{y=0..3}{\mathrm{\Σ}}(x,y)-\mathrm{Left}(y+\mathrm{ModeOffset}\left(\mathrm{mode}\right))$

Wherein, x=0..3, (x y) is the coded image gray value to S, and ModeOffset (mode) is the side-play amount of corresponding different mode, and Left (x) is the image prediction value of prediction piece left side one row.Different mode (Mode) and corresponding modeOffset[to different x values] be defined as follows (be example with the 4x4 piece).

x	Mode	ModeOffset(mode)
			0	18,26,34	-1,0,1
1	18,22,26,30,34	-2,-1,0,1,2
			2	18,20.67,23.33,26,28.67,31.33,34	-3,-2,-1,0,1,2,3
3	18,20,22,24,26,28,30,32,34	-4,-3,-2,-1,0,1,2,3,4

3.2 (mode, x) interpolation is estimated the sad value of each longitudinal edge associative mode 19 to 34 with Int_SAD.This sentences linear interpolation is the sad value that example is calculated each pattern.

SAD_Interploted(mode,x)=(Int_SAD(mode_0,x)*(mode_1-mode)+Int_SAD(mode_1,x)*(mode-mode_0))/(mode_1-mode0),mode=19..34

Wherein mode_0 be in the last table in the corresponding x value less than the Mode value that equals current calculating mode, mode_1 be in the last table in the corresponding x value greater than the Mode value of the current mode of calculating.

3.3 the sad value that computation schema is total;

$\mathrm{SAD}\left(\mathrm{mode}\right)=\underset{x=0...3}{\mathrm{\Σ}}\mathrm{SAD}\_\mathrm{Interploted}(\mathrm{mode},x),\mathrm{mode}=19..34$

4. the RD cost of each pattern is calculated,

Cost(mode)=SAD(mode)+estimate_bit_count(mode);Mode=0..34

Therefrom select three Cost () minimum pattern gather as the candidate frame inner estimation mode.Wherein, estimate_bit_count (mode) adopts the bit number of gained compressed bit stream behind this pattern-coding for this piece image, available does rough estimation by in the context model of current encoder this predictive mode (being mode) being carried out the bit number that variable-length code table (VLC) encodes herein.

Accompanying drawing 5 shows the HEVC encoder block diagram according to the embodiment of the invention.Described HEVC encoder is used for realizing preceding method and technical scheme of the present invention, comprises: motion estimation module, be used for carrying out estimation according to reference frame and present frame, and estimate the error between present frame and the reference frame; Motion compensating module is used for the result according to estimation, and reference frame is carried out motion compensation, and the result of motion compensation is used for interframe encode; Infra-frame prediction is selected module, is used for intra-frame encoding mode is predicted selection, and the prediction selection result sends to intra-framed prediction module; Intra-framed prediction module is used for intra-frame encoding mode is predicted, and prediction result output is used for intraframe coding; All the other modules are module commonly used, respectively as shown in Figure 5.

Described infra-frame prediction selects module and intra-framed prediction module to be respectively applied to realize aforementioned intra-frame encoding mode forecast method step.Wherein, infra-frame prediction selects module to can be used for realizing: utilize the coding result on top or the left side as predicting the outcome pixel value, directly obtain the prediction error value of the integer pixel of every row or column with image block computing to be encoded; The angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance; Integer pixel prediction error value and corresponding angle value thereof by every row or column come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition; The prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode.

Intra-framed prediction module can be used for realizing: the predicated error conduct under the different mode that estimates is assessed the coding cost of different mode with this prediction on intra-frame prediction mode error, thereby finds out optimum predictive mode or find out several predictive modes than top grade.

Each module in the above-mentioned HEVC encoder also can be used for realizing respectively the corresponding steps in the preceding method.

More than describing only is a kind of implementation of this invention.This invention comprises other scheme of finishing by this thought.

It should be appreciated by those skilled in the art, the mode that method and apparatus of the present invention can adopt hardware, software or hardware and software to combine realizes by variety of ways such as microprocessor, digital signal processor, field programmable logic unit or gate arrays.

In sum, though the present invention with the preferred embodiment disclosure as above, yet it is not in order to limit the present invention.The general technical staff of the technical field of the invention without departing from the spirit and scope of the present invention, can do various changes and modification.Therefore, protection scope of the present invention is as the criterion when looking the scope that accompanying Claim defines.

Claims (6)

1. one kind is used for efficient video coding (HEVC) intra-frame encoding mode forecast method, it is characterized in that, comprises following steps:

The first step utilizes the coding result on top or the left side as predicting the outcome pixel value, directly obtains the prediction error value of the integer pixel of every row or column with image block computing to be encoded;

Second step, the angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance;

In the 3rd step, come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition by the integer pixel prediction error value of every row or column and the angle value of correspondence thereof;

In the 4th step, the prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode;

In the 5th step, the predicated error under the different mode that estimates with the method is assessed the coding cost of different mode as this prediction on intra-frame prediction mode error, thereby finds out optimum predictive mode or find out several more excellent predictive modes.

2. according to being used for HEVC intra-frame encoding mode forecast method described in the claim 1, it is characterized in that:

In the first step, directly utilize the direct conduct of coding result on top or the left side to predict the outcome, rather than adopt filtered coding result as predicting the outcome.

3. according to being used for HEVC intra-frame encoding mode forecast method described in the claim 1, it is characterized in that:

In the 4th step, the predicated error of whole predictive mode calculated break the whole up into parts, merge predicated error as the pattern of correspondence by the predicated error with every row or column correspondence.

4. one kind is used for the device that efficient video coding (HEVC) intra-frame encoding mode is predicted, comprising:

Infra-frame prediction is selected module, is used for intra-frame encoding mode is predicted selection, and the prediction selection result sends to intra-framed prediction module;

Intra-framed prediction module is used for intra-frame encoding mode is predicted, and prediction result output is used for the predictive mode that infra-frame prediction selection module is selected the best.

5. as the device that is used for the prediction of efficient video coding (HEVC) intra-frame encoding mode in the claim 4, it is characterized in that described infra-frame prediction selects module also to be used for realizing:

Utilize the coding result on top or the left side as predicting the outcome pixel value, directly obtain the prediction error value of the integer pixel of every row or column with image block computing to be encoded;

The angle of the integer pixel prediction error value correspondence of the good every row or column of calculated in advance;

Integer pixel prediction error value and corresponding angle value thereof by every row or column come interpolation calculation to go out the prediction error value of the pattern of HEVC normalized definition;

The prediction error value that the angle mode of horizontal multirow or vertical multiple row is identical merges the prediction error value as associative mode.

6. as the device that is used for the prediction of efficient video coding (HEVC) intra-frame encoding mode in the claim 5, it is characterized in that described intra-framed prediction module also is used for realizing:

Predicated error conduct under the different mode that estimates is assessed the coding cost of different mode with this prediction on intra-frame prediction mode error, thereby finds out optimum predictive mode or find out several predictive modes than top grade.

Images